Re: [Qemu-devel] [PATCH] pci: Set err to errp directly rather than through error_porpagate()

[Mao Zhongyi]

Hi, Markus

On 05/26/2017 05:43 PM, Igor Mammedov wrote:
> On Fri, 26 May 2017 16:29:25 +0800
> Mao Zhongyi <address@hidden> wrote:
>
> > ioh3420_interrupts_init() and its callers, rp_realize() and
> > pci_qdev_realize() fill error message to local_err, then
> > propagate it to errp by error_porpagate(), which's not necessary.
> > So eliminate it and pass errp directly instead of local_err.
> > Of course, error_propagate() also will be removed.
> >
> > Signed-off-by: Mao Zhongyi <address@hidden>
> > ---
> [...]
>
> > diff --git a/hw/pci/pci.c b/hw/pci/pci.c
> > index 98ccc27..0d4a3ff 100644
> > --- a/hw/pci/pci.c
> > +++ b/hw/pci/pci.c
> > @@ -1982,7 +1982,6 @@ static void pci_qdev_realize(DeviceState *qdev, Error 
> > **errp)
> >  {
> >      PCIDevice *pci_dev = (PCIDevice *)qdev;
> >      PCIDeviceClass *pc = PCI_DEVICE_GET_CLASS(pci_dev);
> > -    Error *local_err = NULL;
> >      PCIBus *bus;
> >      bool is_default_rom;
> >
> > @@ -1999,9 +1998,8 @@ static void pci_qdev_realize(DeviceState *qdev, Error 
> > **errp)
> >          return;
> >
> >      if (pc->realize) {
> > -        pc->realize(pci_dev, &local_err);
> > -        if (local_err) {
> > -            error_propagate(errp, local_err);
> > +        pc->realize(pci_dev, errp);
> > +        if (errp && *errp) {
> >              do_pci_unregister_device(pci_dev);
> >              return;
> >          }
> > @@ -2014,9 +2012,8 @@ static void pci_qdev_realize(DeviceState *qdev, Error 
> > **errp)
> >          is_default_rom = true;
> >      }
> >
> > -    pci_add_option_rom(pci_dev, is_default_rom, &local_err);
> > -    if (local_err) {
> > -        error_propagate(errp, local_err);
> > +    pci_add_option_rom(pci_dev, is_default_rom, errp);
> > +    if (errp && *errp) {
> >          pci_qdev_unrealize(DEVICE(pci_dev), NULL);
> >          return;
> >      }
>
> dropping local_error here looks wrong since it's used
> to check error status inside these functions and to undo
> side effects in case of failure.
>
> consider if caller pass errp = NULL
> then error handling path won't be executed.

Thanks for you detailed explanation, I ignored this question
actually. The specified error message will not be converted to
Error. Because error handling function returned directly when
caller pass errp = NULL, right?


> So keep the rest of the patch but drop above hunks.

I will drop this part of wrong process and commit the next
version right away.

Thanks a lot.
Mao